Item

Abstract

We model how opponent interactions between 5HT and DA in the vertebrate midbrain and striatum may influence the learning of predictions of rewards and punishments in classical conditioning and the learning of actions in operant conditioning. Opponent interactions may be both direct, with 5HT repectors in the VTA and substantia nigra and DA receptors in the raphe nuclei, and indirect, at common target sites in the striatum and amygdala. Examples of the interactions between these two systems include the 5HT-sensitivity of amphetamine-potentiated operant behaviors. The model takes explicit account of the involvement of DA in appetitive behaviors and stimulus-specific orienting and approach behaviors and 5HT in aversive and stimulus-specific defensive behaviors. It is also consistent with observations of increases and decreases from a (low) baseline of the activity of DA projecting cells in the VTA and substantia nigra in the face of salient and action-provoking stimuli that are not associated with the delivery of reward. These DA activity changes are putatively independent of 5HT. The model is an extension of temporal difference models (also called prediction-error or actor-critic models) of the DA system. Prediction learning is associated with the basolateral amygdala and ventral striatum; action learning is associated with the dorsal striatum.